Tracking Possibly Unresolved Targets with PMHT

Abstract

In tracking scenarios with high target density, it is possible that two (or more) targets may be close enough to be unresolved by the sensor. In this case, the unresolved targets produce only a single measurement, and this may cause the tracking algorithm to incorrectly terminate track on one of the targets. This paper presents an extension of the probabilistic multi-hypothesis tracker (PMHT) to allow for the possibility of unresolved targets. The extended algorithm is compared with the standard PMHT, and with the joint probabilistic data association algorithm for unresolved measurements.